Learn by changing the model

Interactive machine learning playgrounds

Move points, change parameters, and see the algorithm update immediately. Every playground runs locally with JavaScript and links the visualization back to a reusable API.

No setupRuns in your browserBuilt for teaching and exploration

PCA visualizationPaste multidimensional data, inspect explained variance, and rotate a valid 2D loading basis.KNN visualizationDraw labeled points, move a query, and compare k, distance metrics, and weighted voting.Gradient descent visualizationCompare SGD, Momentum, and Adam trajectories on convex and non-convex objectives.K-Means visualizationStep through assignment and centroid updates on blobs, moons, and uneven clusters.Decision tree visualizationTune a classifier and connect its decision regions to the fitted rule hierarchy.Random Forest visualizationCompare a single tree with a bootstrap-aggregated forest on noisy data.

Browser-native algorithm visualizations

These playgrounds turn the JavaScript implementations in @kanaries/ml into inspectable learning tools. They are designed for students, teachers, frontend developers, and data practitioners who want to test an intuition without opening a notebook or sending data to a server.

Start with PCA to understand dimensionality reduction, KNN to see local voting, gradient descent to compare optimizers, or K-Means to follow assignment and centroid updates. Continue to the ML calculators when you need a numerical result rather than an algorithm demonstration.